Abstract. Multiple states of woody cover under similar climate conditions are found in both conceptual models and observations. Due to the limitation of the observed woody cover data set, it is unclear whether the observed bimodality is caused by the presence of multiple stable states or is due to dynamic growth processes of vegetation. In this study, we combine a woody cover data set with an aboveground biomass data set to investigate the simultaneous occurrences of savanna and forest states under the same precipitation forcing. To interpret the results we use a recently developed vegetation dynamics model (the Balanced Optimality Structure Vegetation Model), in which the effect of fires is included. Our results show that bimodality also exists in aboveground biomass and retrieved vegetation structure. In addition, the observed savanna distribution can be understood as derived from a stable state and a slightly drifting (transient) state, the latter having the potential to shift to the forest state. Finally, the results indicate that vegetation structure (horizontal vs. vertical leaf extent) is a crucial component for the existence of bimodality.